New features: * Add possibility to build a P(X,Y) model in two steps, i.e. a specific model for forecasting where a difference is made between internal links X-X or Y-Y and cross links X-Y. Ny is the number of variables Y out of Nv= NvX+NvY. k_XY is the target connectivity contribution of links X-X and Y-Y before switching to X-Y. * Add new parameter links_selection for pre-selecting a short list of possible links to chose in for building the model. * when the requested connectivity is attained, the couplings are output in file `couplings.dat'. * New options -allcouplings added to output also the intermediary files. Updates: * Code optimization: due to code re-ordering and auto-vectorization by the compiler, execution is now 2.85 times faster on an example with 2740 variables. * better formatting and organization of the options in -h output.